EconomiX http://economix.fr
Document de Travail Working Paper 2014-36
Does nominal rigidity mislead our perception of the exchange rate pass-through?
Olivier de Bandt Tovonony Razafindrabe
Université de Paris Ouest Nanterre La Défense (bâtiment G) 200, Avenue de la République 92001 NANTERRE CEDEX
UMR 7235
Tél et Fax : 33.(0)1.40.97.59.07 Email :
[email protected]
Does nominal rigidity mislead our perception of the exchange rate pass-through? Olivier de Bandty
Tovonony Raza…ndrabez
We owe a special thanks to the French National Institute of Statistics and Economic Studies (INSEE) for welcoming us during a few months to access to their database and providing us unpublished data of French import prices. Indeed, we owe a huge debt of gratitude to Alain Gallais and Charles Pilarski for their precisous assistance during the data extraction. Comments by participants in the European Economic Association 2013 Meeting, as well as the French Economic Association 2013 Annual Meeting are gratefully acknowledged. The views expressed in this paper are those of the authors and do not necessarily express the views of the INSEE or the ACPR. y Banque de France - ACPR and EconomiX-CNRS, Université Paris Ouest Nanterre la Défense. z EconomiX-CNRS, Université Paris Ouest Nanterre la Défense.
1
Abstract Relying on a novel dataset of detailed micro-data on import prices, this paper explores the close link that exists between nominal import price rigidity and the extent of exchange rate pass-through (ERPT). We show that previous evidence in favor of incomplete and low value of ERPT in the empirical literature may be explained by two factors: the relative importance of small variations in the exchange rate and, mainly, nominal rigidity. Once nominal rigidity is taken into account, we …nd for French manufacturers that ERPT may be incomplete in the short run, but with relatively high value, and complete in the long run. In addition, assessing non-linearity and asymmetry issues, we provide evidence that the shape of the import price reaction function is distorted by the presence of nominal rigidity. Indeed, the linearity assumption is veri…ed once nominal rigidity is taken into account. However, in the case where it is rejected, the import price reaction function is concave rather than convex, indicating that …rms aim at protecting market shares. As a consequence, the common belief that "prices rise faster than they fall" is the results of nominal import price rigidity as far as ERPT is concerned.
Keywords: Exchange rate pass-through, nominal rigidity, import price JEL: F31, E31, C23
2
1
Introduction
The phenomenon of exchange rate pass-through, henceforth ERPT, is an issue of a particular relevance for policy makers and has received a great deal of attention in both the theoretical and empirical literature. As shown by Obstfeld and Rogo¤ (2000), Betts and Devereux (2000) and Engel (2003), the traditional Keynesian e¤ects of exchange rate movements such as expenditure switching e¤ect, consumption risk sharing, terms of trade e¤ect, international competitiveness, welfare analysis and the law of one price assumption are altered by the extent of the ERPT. In addition, the ERPT plays an important role for monetary policy analysis as evidenced by Gagnon and Ihrig (2001) and Monacelli (2005). Thus, there is no doubt concerning the impact of the ERPT on macro-economic environment and its important role for policy decision making. However, no clear consensus has been reached so far concerning its extent. From a positive perspective, the extent of the ERPT is low and incomplete as a result of an increasing degree of pricing-to-market behavior of …rms or of a change in the import consumption basket towards products characterized by a high degree of pricing-to-market.1 From a normative perspective, the pass-through is incomplete in the short run due to price stickiness but complete in the long run.2 The aim of this paper is not to reconcile these two di¤erent points of view. However, it highlights the important role of nominal rigidity in explaining the commonly observed phenomenon of incomplete ERPT on the basis of a unique dataset of import prices recorded at company level. As argued by Burstein and Gopinath (2013), the latter phenomenon might be explained by two factors: nominal rigidity and pricing-behavior of …rms through a mark-up adjustment. This paper focuses on the normative approach by studying the close link that exists between price stickiness and ERPT. The reason why nominal rigidity should be the main polar star in this study when trying to rationalize incomplete ERPT is twofolds. First, the use of micro-data at company level o¤ers the possibility of controlling for nominal rigidity by conditioning import prices on adjustment, i.e. to take into account that import prices adjustments take place at discrete points in time while exchange rates move continuously. It is worth noting that data availability for Euro-area import prices prevented so far previous studies from using thoses indices for ERPT analysis.3 Second, given that the literature that 1 We can refer to the seminal work of Betts and Devereux (2000) for theoretical approach of this literature. One can cite, however, among the large number of empirical studies in this …eld, the work of Marazzi and Sheets (2007), Bouakez and Rebei (2008), Gust et al. (2010), and more recently Auer and Schoenle (2013). 2 This strand of the literature is composed in a large part by the New-Keynesian general equilibrium model such as Obstfeld and Rogo¤ (2000), Smets and Wouters (2002) and Monacelli (2005). To cite a few empirical studies that support this approach, we can refer to the work of Campa and Goldberg (2002), Campa et al. (2005) and Campa and Minguez (2006). 3 The import price index was o¢ cially introduced by the Regulation (EC) No. 1158/2005 of the European Parliament and of the Council of 6 July 2005 amending Council Regulation (EC) No. 1165/98 concerning short-term statistics, and was published in the O¢ cial Journal of 22 July 2005. Among the scarce recent studies that use …rm-product-level imports prices data for the Euro-area, one can cite Amiti et al. (2012) who study the role of maket share and/or import intensity of Belgian exporting …rms in explaining the phenomenon of incomplete ERPT.
3
supports the positive approach generally shares the same common point of using data that exhibit some degree of nominal rigidity, it is legitimate to ask whether these …ndings are distorted by the presence of price stickiness. Therefore, isolating the role of nominal rigidity in explaining the phenomenon of incomplete ERPT will help us disentangle about the two alternative explanations advanced in the literature. To tackle this issue, we use unpublished micro-data of French import prices at the monthly frequency for the period 2005M8 to 2011M5 made available to us by the French National Statistical Agency (INSEE). First, relying on statistical analysis commonly used in the consumer and producer prices micro-data studies, we advance some new import prices stylized facts in order to get some insight of the close link that exists between nominal rigidity and ERPT. Namely, we show that import prices are more prone to frequent changes and that the amplitude of these price changes is relatively small. Second, we …nd that evidence in favor of incomplete and low values of ERPT is the result of nominal import price rigidity and small exchange rate changes that occur more frequently. Once price stickiness is taken into account, the pass-through is high, if not complete in the short run, and complete in the long run. Third, we provide evidence that price stickiness distorts the shape of the import price reaction function to exchange rate changes. That is, as far as ERPT is concerned, the common belief that "prices rise faster than they fall" (Pelzman (2000)), which is formally characterized by a convex import price reaction function, is the result of nominal import price rigidity. Once nominal rigidity is taken into account, the linearity assumption is veri…ed. But in some cases where the latter is rejected, import price reaction function is concave, indicating that …rms aim at protecting market share. In light on these …ndings, we contribute to the literature by proposing some new stylized facts on import prices and by contributing to shed light on the blurred but important issue of the extent of ERPT. Our results mainly show that nominal rigidity plays an important role in explaining the phenomenon of incomplete ERPT and that its presence in the import prices misleads our perception of the observed ERPT, especially when using aggregate prices. The rest of the paper is organized as follows: section 2 describes the data; section 3 proposes some import price related facts; section 4 assesses the extent of ERPT; Section 5 concludes.
2
Data description
Data used in this study are based on the unpublished monthly import price index issued by the French National Statistical Agency (INSEE).4 The monthly import prices provided by the INSEE are "at the dock" company-level transaction prices of French importing …rms for the month considered. This is the …nest possible degree of import prices that can be recorded for France. It corresponds to the 6-digit level which we denote as the entry level items, 4
We are thankful to the INSEE for making these data available to us under con…dentiality restriction following the agreement of the Secrecy Committee (Comité du Secret Statistique).
4
henceforth ELI. This amounts to a set of around 4150 individual prices each month. Notice that, in contrast with some countries such as the United States and the United Kingdom where prices are collected on a particular day at the beginning of the month, …rms give their transaction price for the month. This methodology could lead to collect what is called a "mean price" but in turn would avoid collecting a "list price" (without rebate and discount) which may have a di¤erent evolution than the actual transaction price. Therefore, given the methodology adopted by the INSEE, the exchange rate used in this study is the monthly averaged one. Gopinath and Rigobon (2008) used the end of the month exchange rate but lagged one period due to the price collection procedure. Our dataset covers the 2005M8 to 2011M5 period. In order to present and compare our results to other studies, we rely on some product classi…cations - or sector/product break downs - generally used in this type of literature. We refer to the classi…cation used by Rauch (1999), Vermeulen et al. (2012) and the 2-digit European Statistical Classi…cation of Economic Activities, henceforth NACE2.5 Rauch (1999) classi…ed goods into di¤erentiated products (DP), homogeneous products (HG) that are traded on an exchange and reference priced (RP) products whose prices are listed in a trade publication. Vermeulen et al. (2012) classi…ed goods into six categories that are consumer food products (CFP), consumer non-food non-durables (CNFND), consumer durables (CD), intermediate goods (IG), energy (ENER) and capital goods (CG). Moreover, we choose to use the NACE2 classi…cation, because it is the level at which import prices begin to be published by INSEE and Eurostat. Some de…nitions are necessary for the rest of the paper. They are standard de…nitions in the literature on price persistence. Quote-line (also refered to as price quote in the literature) indicates the time, de…ned in terms of number of months in our study, that elapses between the entry of a given ELI in the sample and its exit. Table 1 in the appendix D depicts quotelines for the di¤erent classi…cations listed above. In general, the weighted (respectively, the non-weighted) mean of quote-lines amounts to 38:23 months (respectively, 39:70). That is, import prices at the entry level last in general three years in the sample. However, quote-lines are heterogeneous across products. Some categories of products rotate more frequently, such as the manufacture of leather and related products (LEATH) that rotate in average every two years, while some products last longer such as the manufacture of beverages that rotates in average every …ve years. Price spell refers to the time that elapses between the current observed price change and the last one. There may be several price spells, possibly with di¤erent durations which henceforth are denoted as s, within a given quote-line. Moreover, price spells are heterogeneous across products. There are products that have frequent price changes, and thus a low price spell, as well as products that do not have any price change during their entire life in the sample. Therefore, some assumptions are sometimes needed regarding price change concerning the entry and exit of products. We consider that a price introduction also corresponds to a price change. That is, we do not take left censoring into account. Indeed, the value 5
Classi…cations and tables of correspondence are presented in a supplement available upon request.
5
of the import price prior to the introduction of the ELI in the sample is unknown. In turn, several speci…cations are possible concerning exit. It will be indicated throughout the analysis whether or not we consider right censoring, which we will refer to as censoring for short.6 Finally, the weight used throughout this study is calculated from individual import turnovers that were also provided to us alongside the import prices. However, those weights, denoted wi , are only available until the 4-digit classi…cation. Hence, we use the turnover to compute the weighted mean of the variables of interest until 4-digit and from there on, compute an arithmetic mean until a given classi…cation provided above.
3
Some stylized facts on import prices
In order to guide us for the estimation of the ERPT, it is useful to contemplate a few summary statistics of our database and derive a few stylised facts. The use of micro-data in the literature of the ERPT has received a great deal of interest since the seminal work of Gopinath and Rigobon (2008), henceforth G.R. For instance, Gopinath, Itskhoki and Rigobon (2010) estimate the extent of ERPT depending on the currency (Dollar or non-Dollar) in which import prices are denominated. In turn, Gopinath and Itskhoki (2010) focus on the relationship between the frequency of price changes and the extent of ERPT. For the Euro-area, while there are many studies conducted by the In‡ation Persistence Network team at the European Central Bank that use micro-data, they focus only on consumer and producer price indexes. Álvarez et al. (2005) provide a comprehensive review of this literature. To the best of our best knowledge, our study is the …rst to provide detailed statistics on import prices at the micro-level and their use for the ERPT analysis for the Euro-area and mainly for France. This lack in the literature may be explained by data availability constraints, given that the import price index exists o¢ cially for the Euro-area since 2005. It is thus of interest to present some important import price statistical facts in order to complement the literature and to better understand some features of the observed ERPT.
3.1
Hazard rate
Following the seminal paper of Monacelli (2005), a signi…cant part of the literature relates the extent of ERPT to the degree of nominal price rigidity. The latter is de…ned as the probability that import prices remain unchanged for a given period.7 For micro-data based studies such 6
Nakamura and Steinsson (2012) and Gagnon et al. (2012) go in depth into this problem of missing price change before entry and after exit, which they call "product replacement bias". The …rst authors argue that this generates a large downward bias in the estimation of the ERPT using aggregate import price index and …nd an estimate of the long run pass-through that amounts to 0:60 after correcting the bias, whereas the second authors argue that this downward bias is mitigated and …nd in turn an estimate of 0:28 after the bias correction. This issue is, however, beyond the scope of this study. 7 It is common to measure this rigidity by the means of Calvo parameter in the open macro-economic models, namely for the dynamic stochastic general equilibrium model used extensively for policy analysis as in Adolfson et al. (2007) and Christo¤el et al. (2008). Moreover, it is generally assumed that the degree of nominal rigidity
6
as G.R, it is common to measure nominal rigidity by the means of the hazard rate. This is the instantaneous probability of a price change (of leaving a state) conditional on no price change (on survival) until the considered period. In the following, we assess hazard rate conditional on price duration or price spell. i Let us de…ne pm it the import price index of a given quote-line i for the period t, It is a i;fs= g m price change indicator that takes the value of 1 if pm it 6= pit 1 and 0 otherwise, and It a price spell indicator that takes the value of 1 if the price spell s lasts months and 0 otherwise. Rather than specifying a given functional form, as the exponential function that gives a constant hazard rate,8 we use the speci…cation adopted in Klenow and Kryvtsov (2008), henceforth K.K, and compute the hazard rate conditional on price spell. That is,9 "P # i i;fs= g X X t It :It j fs = g = wi i j fs = g = wi P i;fs= g (1) I i i t t Recall that wi is the weight of the individual product i. This yields the following Figure 1 of the hazard rate by spell, that is, the probability of price change conditional on the length of the period (in months) represented in the horizontal axis during which prices remain …xed. Figure 1: Hazard rate conditional on price spell (from 1 to 24 months)
As shown in Figure 1, the hazard rate of the …rst spell, namely the probability of a price change every month, is slightly less than 0.6. It is three times greater than the hazard rate of the second and third spells which indicate that when prices have not changed for 2 or 3 is constant and does not depend on the number of periods elapsed since the last price change. 8 See for example G.R and Klenow and Kryvtsov (2008). The use of the exponential function assume ex-ante that the hazard rate is constant. We opt for non-parametric speci…cation and assess this hypothesis. 9 Detailed and intuitive derivation of the following conditional hazard rate can be found in the Appendix A.1.
7
months, the probability of a price change is around 0.2. This …nding is mainly explained by the volatility of the exchange rate that contributes to frequent changes (lower degree of stickiness) of the import prices compared to the producer and consumer prices. It is worth noting that 57:6% of the observations fall in the …rst three spells. Moreover, there is clearly a decline in the hazard rate as the length of the spell increases. That is, the stickier the prices, the lower the probability of price changes. However, as is common in duration analysis, this situation may occur as a result of an aggregation of heterogeneous individuals. That is, the hazard rate may e¤ectively be constant within a category of goods but may di¤er across di¤erent categories even if they are driven by the same covariates due to unobserved heterogeneity.10 Aggregating across heterogenous categories of goods which have constant but di¤erent hazard rates, in other words ignoring unobserved heterogeneity, will lead to a decreasing shape of the hazard rate. Therefore, before drawing any conclusion, it is necessary to correct the conditional hazard rate above for a possible unobserved heterogeneity by normalizing hazard rates within di¤erent categories of goods before aggregation. To this end, let i be the unconditional hazard rate, which is the frequency of price changes, of a given import price i. In order to obtain normalization coe¢ cients which we refer to as …xed e¤ects, let us sort import prices into decile based on the value of i and compute the mean of the unconditional hazard d for each decile. Therefore, the corrected hazard rate (before aggregation) is obtained by dividing the conditional hazard rate i j fs = g of the ELI i de…ned in (1) by the corresponding average frequency of price changes d of the decile category to which the individual import price belongs, which we denote di for the sake of clarity. Therefore, the aggregate corrected hazard rate with decile …xed e¤ects is given by: ~ j fs = g =
X
wi
i
i j fs
= g
d i
(2)
This yields the corrected hazard rate by spell dispayed in Figure 2. 10
As far as exchange rate pass-through is concerned, heterogeneity may stem from micro and macro-factors. Devereux and al. (2004) and Goldberg and Tille (2008) study the explicit role of industry-speci…c features in contrast to macro-economic variability. They argue that the macro-economic environment plays an important role in the extent of pass-through for di¤erentiated products. In contrast, producers of homogeneous goods characterized by high elasticity of demand, the so-called Walrasian goods, are much more in‡uenced by microdeterminants.
8
Figure 2: Corrected hazard rate conditional on price spell (from 1 to 24 months)
Clearly, the decreasing shape of the hazard rate disappears once we take into account the unobserved heterogeneity, but the fact that prices in the …rst spell have a higher hazard rate remains.11 Thus, the probability of import price changes is higher within the …rst spell relative to other spells. In contrast to …ndings of K.K that concerns PPI and CPI, this result indicates that import prices are more prone to frequent changes due to exchange rate volatility. Finally, apart from some spikes every twelve months, the hazard rate is essentially ‡at. Recall that the number of prices that last one and two years is small, more precisely 1:4% and 0:5% of the total observations, so the higher value of the hazard rate for those spells does not really matter.
3.2
Import prices change more frequently
As shown in Table 1, the average duration of the import price stickiness12 amounts to 6:67 months. That is, changes in the import price index occur in average twice a year. This duration is short as compared to a change of once a year for the Euro-area CPI (13:0 months) and PPI (10:8 months) in Álvarez et al. (2006) and the import price duration of 10:6 months for the United States in G.R.13 11
Taking into account only complete spell, that is, not considering the introduction of a given quote-line in the sample as a price change, does not alter the conclusion. 12 As is standard in the literature of Calvo price setting, duration is calculated as the inverse of the unconditional probability (frequency) of price changes. It is given by 1=(1 ) where denotes the (Calvo) probability that …rms keep price unchanged for a given period that is independent of the time elapsed since …rms last reset price. 13 Compared to the results of G.R., di¤erence in the value of import price duration may stems from various factors such as the di¤erence in import basket composition, the data collection procedure and the di¤erence in the degree of entry level items, namely, G.R. used United States data at 10-digit level compared to the French data at 6-digit we use in this study. This di¤erence was also pointed out by Berman et al. (2012).
9
However, price rigidity is heterogeneous across products. As one may expect, homogeneous goods (HOMO) have a shorter price duration which is only equal to 1:65 month. Indeed, this category of goods, which is traded essentially on an exchange, is characterized by a high price elasticity of demand. It is more sensitive to international price ‡uctuations within raw materials market. For instance, the manufacture of electricity, gas, stream and air conditioning supply (EGSA) has its price changed each month, that of petroleum products (PETRO) and mining and quarrying (MINQUA) have their prices changed every 2 months (respectively, 2:29 and 1:66 months). On the contrary, di¤erentiated products (DP) are characterized by a low price elasticity of demand, at least in the short run. Hence, they are characterized by a higher price stickiness which amounts to 7:44 months. For instance, the manufacture of furniture (FURN) has the highest price duration, reaching 15 months.
3.3
The amplitude of import price changes is small
In our sample, the weighted (respectively, the non-weighted) mean of the absolute size of the import price changes amounts to 4:26% (respectively, 4:32%). This is relalively small compared to the value of 8:2% found by G.R for the United States. However, this should be related to our earlier …ndings that prices change relatively more often than in other countries. When splitting into price increases and decreases, the weighted average size amounts respectively to 5:9% and 3:4%. These are small compared to the value of 8:2% and 10% found by Álvarez et al. (2006) for the consumer price index of the Euro-area. Note that the inter-quartile interval of the size of import price changes ranges from 0:56% to 1:26%. A value that is relatively small. Figure 3: Evolution of the absolute size of price changes in percentage
Plots of the evolution of the absolute size of price changes in Figure 3 show that apart 10
from a sharp increase in the last quarter of 2008 during the world trade collapse and which was associated to an appreciation of the US Dollar, it remains small and ‡uctuates around 3% before 2008 and around 4% thereafter.
4
Exchange rate pass-through
On the basis of these stylized facts which show that import prices are more volatile than in the US, but exhibit a signi…cant degree of rigidity, we now estimate the degree of the ERPT. First, to serve as a benchmark and in order to compare our results to those of previous studies, we proceed to panel estimation of the ERPT without any transformation of the data. Second, we take into account the presence of nominal rigidity in the estimation of the ERPT. For that purpose, we condition our data on price adjustments and price changes.14 Intuitively, conditioning data on price adjustment permits to measure, for each complete spell, the extent of price adjustment that is explained by the cumulative sum of exchange rate changes during the spell. While conditioning data on price change considers, for each period, only products that have a price change. This distinction is important in order to assess if …rms take into account either the cumulative sum over the entire price spell or the current, and possibly lags, of exchange rate changes whenever they adjust their prices. Third, we assess non-linearity and asymmetry issues using the threshold model under the panel and conditional on price adjustment speci…cations. It permits to assess how common results of the ERPT are altered by the presence of nominal rigidity as far as non-linearity and asymmetry issues are concerned.
4.1
Panel estimation
An important puzzle in the empirical literature of the ERPT is the choice of partner’s export prices. This includes both the relevant exchange rate and a proxy for domestic costs. Thus, we adopt the same approach as Nakamura and Steinsson (2012) and use the real e¤ective exchange rate due to the lack of information on the country of origin and, hence, on the relevant nominal bilateral exchange rate to use. This also has the advantage of taking into account the change in the relative price (PPI) between home and foreign directly in the real exchange rate, rendering the equation of pass-through homogeneous across countries and 14
Details of these data transformation can be found in the Appendix B.1.
11
products.15 Fixed e¤ect panel estimates of the ERPT are given by the following equation: pm t
pppi = t
6 X
j
qt
j
+ d0w + "t
(3)
j=0
ppi where pm denote respectively the (log) stacked import and domestic producer prices, t and pt qt is the logarithm of the e¤ective real exchange rate,16 and dw is the vector of dummy variables corresponding to the 4-digit level of the NACE2 classi…cation used to control individual heterogeneity. Results are presented in Table 2 of appendix D. It is worth noting that for the di¤erent speci…cations adopted in this paper, our results are robust to the number of lags chosen in the estimation as shown in appendix C. Pass-through is incomplete both in the short and long run with estimates that amount respectively to 0:214 and 0:475. G.R obtained similar estimates where the weighted average of the ERPT for market transaction amounts to 0:21 and is in general around 0:33 or lower across individual products. For the Rauch classi…cation, pass-through into import prices of di¤erentiated products is lower and incomplete both at the short and long run, respectively 0:196 and 0:399. Pass-through in turn is low in the short run, 0:186 and 0:281, but high in the long run, 0:722 and 0:700, respectively for homogeneous and reference priced products. These are common results in the pass-through empirical literature. Di¤erentiated products that are characterized by high transformation have lower price-elasticity of demand. In turn, homogeneous and reference priced products are more sensitive to international price ‡uctuations within raw materials market and are characterized by high price-elasticity of demand. Therefore, ERPT is higher for these categories of products compared to that of di¤erentiated products. At a more disaggregated level, there is complete pass-through, in both the short and long run, for the mining and quarrying (MINQUA), the manufacture of coke and re…ned petroleum (PETRO) and the manufacture of electrical equipment (ELEC) products.
4.2
Stickiness and pass-through
If we rely on the panel speci…cation above, one could conclude, as is common in the literature, that pass-through is in general low and incomplete at the short run and it remains true for a large number of products at the long run, especially for di¤erentiated products. From a 15
Let us start from the following simple equation of the import price with full and immediate pass-through: x; = et + px; is proxied by exporter’s producer price pt ;ppi and t . Assuming that the foreign export price pt subtracting importer’s producer price pppi on both sides of the equation yield pm pppi = et +pt ;ppi pppi t t t t . The right-hand side variable is nothing but the real e¤ective exchange rate qt . Taking …rst di¤erence and adding a constant, an error term and lags to real e¤ective exchange rate, yield the equation (3). See Barhoumi (2006) for a discussion. Baillu and Fujii (2004) in turn use unit labor cost based on real e¤ective exchange rate in their study. 16 The e¤ective real exchange rate used in this study is de…ned as a geometric weighted average of the relative producer prices (expressed in the same currency) between the Euro-area and each of its 20 major trading partners. pm t
12
positive perspective, this phenomenon is explained as an increase in the degree of pricingto-market or a declining value of the pass-through. Initiated by the seminal work of Betts and Devereux (2000), this strand of the literature supports the hypothesis that an increasing fraction of …rms practice the Local Currency Pricing - Pricing-to-Market (LCP-PTM) strategy relative to those that practice the Producer Currency Pricing (PCP) one. Thus, as the number of …rms that practice LCP increases, the ERPT becomes low and incomplete both at the short and long run. From a normative perspective, initiated by the seminal paper of Monacceli (2005), short term ERPT might be incomplete as a result of nominal import price rigidity but long term ERPT instead is complete. The degree of price stickiness in turn is measured by the means of the Calvo import price rigidity. Thus, as argued by Burstein and Gopinath (2013), the phenomenon of observed incomplete ERPT is the result of Pricing-to-Market behavior of …rms through mark-up adjustment or/and nominal import price rigidity. One may therefore wonder what happens to the extent of ERPT once nominal price rigidity is taken into account. This is of particular importance in order to disentangle the two alternative explanations advanced in the literature. It will help us to better understand some features of the ERPT, such as the role of pricing-to-market, passthrough dynamics, currency invoicing, measurement errors, etc, without having results that are distorted by nominal rigidity.17 4.2.1
Conditioning on price adjustment
In order to take into account the presence of nominal price stickiness as in Gopinath, Itskhoki and Rigobon (2010), we condition the estimation of the extent of ERPT on import price adjustments. To this end, we run a regression were the dependent variable is de…ned as the di¤erence between the (log) value of the current import price when adjustment is observed and that of the previous price change.18 The elapsed time during price adjustments, the price spell, depends on the degree of price stickiness and is di¤erent across categories of products. By the same analogy, independent variables are de…ned as the (log) cumulative sum during the import price spell. Formally, estimates of the ERPT conditional on price adjustment result from the following regression:19
s
pm t
pppi t
=
c 0 rs qt
+
6 X
c j
qs;t
s j
+ d0w + "t
(4)
j=1
17
This part is not tackled in this paper. Our …rst objective is to estimate the extent of pass-through and to present facts that have to be taken into account to avoid misleading interpretation due to the presence of nominal price rigidity. This is however an important topic for future research. 18 In the case of the …rst price change, it is the di¤erence between the (log) value of the import price where the …rst price change is observed and the one that corresponds to the introduction of the product. For the last spell before the exit, if no price change is observed before the item leaves the sample, this price spell is discarded. That is, we do not consider censoring. 19 See appendix B.2 for detailed explanation of the conditional on price adjustment speci…cation.
13
where the subindex s refers to a complete spell of length s, s is the di¤erence operator over the spell of length or duration s, rs is the cumulative sum operator over the price spell and qs;t s j denotes the (log) change in the real exchange rate j-month prior to the s-th spell. The short run ERPT conditional on price adjustment is measured by the coe¢ cient c whereas that of the long run is given by the sum of the pass-through of the cumulative 0 c ) and the pass-through of the …rst 6 months past exchange rate changes during the spell ( P 0 P exchange rate changes prior to the spell ( 6j=1 cj ), that is, 6j=0 cj . Results that are depicted between square brackets in Table 2 in the appendix show that ERPT is incomplete in the short run, whereas it is complete in the long run. In stark contrast with the estimated values found in the panel speci…cation, the short run estimate is slightly higher and amounts to 0:32 whereas it amounts to 0:86 in the long run and is not statistically di¤erent from one (complete). The same conclusion can be drawn for the Rauch classi…cation. 4.2.2
Conditioning on price change
In the next speci…cation, we condition the estimation of the ERPT on price changes. That is, within each 4-digit category of goods and for each time period, we only consider goods that have a price change. Taking the mean of prices that have changed across individual goods for each time period yields a time series observations for each 4-digit good. Therefore, to estimate the extent of ERPT conditional on price changes, we run the following regression:20 pm t
= pppi t
6 X
j
qt
j
+ d0w + "t
(5)
j=0
where pm t denotes (stacked observation of) the mean of the import price changes across goods within the 4-digit bin at the period t. This speci…cation di¤ers from equation (4) - where changes were conditional on price adjustment - in the sense that it gives now a picture of the pass-through where the nominal rigidity is corrected for each time period rather than for each spell. The correct interpretation is now the following: if …rms change their prices, to what extent the current (not the cumulative) exchange rate changes and its lags, six months in our speci…cation, are passed-through changes in price? Here lags take into account possible delays in the response of import prices to previous exchange rate movements. Table 2 in the appendix reports between parenthesis the pass-through estimates resulting from the conditional on price change speci…cation. The conclusion is relatively the same as that for the conditional on price adjustment speci…cation except that, and this is of a great deal of importance, the short run ERPT is now high, if not complete, with the weighted mean that amounts to 0:60 for a large number of products. That is, the observed low estimates of the ERPT, as found in subsection 4.1, may be explained in large part by the presence of price stickiness. Once the latter is taken into account, pass-through estimates are in general high if not complete both at the short and long run. This …nding, combined with the one obtained in the conditional on price 20
See appendix B.3 for detailed explanation of the conditional on price change speci…cation.
14
adjustment speci…cation, is a support of the staggered price setting fashion à la Calvo or à la Taylor commonly used in the general equilibrium models of the ERPT. A similar conclusion is advanced by K.K.
4.3
Non-linearity and asymmetry
The phenomenon of asymmetry and non-linearity in the ERPT has received a great deal of attention recently. This strand of the literature21 tries to answer the following question: Do an appreciation and a depreciation be passed-through into import prices in the same magnitude (the phenomenon of asymmetry), and does the size of the exchange rate changes in‡uence the magnitude of the pass-through (the phenomenon of non-linearity)? In order to see how results are sensitive to price stickiness, we use two speci…cations which are (1) the pooled panel and (2) the conditional on price adjustment. The …rst speci…cation, which we denote "panel" in Table 3, corresponds to the speci…cation generally used, using micro or macro-data, in the literature of asymmetry and non-linearity and thus will permit to compare our results with those of other studies. The second speci…cation in turn, which we denote "conditional on price adjustment", henceforth "cpa", permits to isolate the phenomenon of asymmetry and non-linearity, if they exist, from nominal price rigidity. Notice that the conditional on price change speci…cation is not retained due to the …nding of Buissière (2013). He argued that aggregation of the import prices leads to non-linearities that cancel each other given the sign of the convexity of the export reaction function that varies across di¤erent trading partners. 4.3.1
Nominal rigidity distorts the shape of the import price reaction function
We assess the asymmetry and non-linearity issues using the Threshold model that is described in the appendix B.4. Positive (respectively, negative) signi…cant estimate of the Threshold parameters in Table 3 indicates a more (respectively, less) than proportional e¤ect of the exchange rate changes into import price changes. Note that for the sake of clarity, only signi…cant estimates are reported and that non-signi…cance of the parameters supports the linearity assumption. For the panel speci…cation, a depreciation has a total impact of 0:33, while an appreciation has a corresponding impact of 0:09 on the import price changes in the short run. This is a support of the convex import price reaction function that relies on two disctinct micro-economic behavioral assumptions. The …rst one concerns the price downward rigidity at the export side during an appreciation of the exporter’s currency. This will yield a higher ERPT at the import side during a depreciation of importer’s currency and is characterized by a positive signi…cant estimate of . The second assumption concerns the quantity upward rigidity on the export side during a depreciation of exporter’s currency. Competitiveness gain due to depreciation leads exporting …rms to raise prices rather than quantities due to production capacity constraints or when it is more costly to raise production compared to the gain issued by the rise in demand. This will yield a lower ERPT at the 21
Bussière (2013) provides a comprehensive review of this literature.
15
import side during an appreciation of importer’s currency and is characterized by a negative signi…cant estimate of . For the cpa speci…cation however, di¤erent but important results emerge. In the short run, the linearity assumption could not be rejected. In the long run in turn, a depreciation has a total impact of 0:65, while an appreciation has a corresponding impact of 1:27 on the import price changes. This is a support of the concave import price reaction function which indicates that …rms are more concerned with demand conditions. In other words, once nominal rigidity is taken into account, linearity assumption prevails, and if it is not the case, as for the long run ERPT, import price reaction function curvature is no longer convex but rather concave.22 Therefore, and as long as ERPT is concerned, the common belief that "Prices rise faster than they fall" in response to exchange rate changes is the result of nominal import price rigidity. Firms are concerned with demand conditions once price stickiness is taken into account. It is worth noting that the convex shape of the import price reaction function remains even if unobserved heterogeneity across products has been taken into account by the means of …xed e¤ects in the panel speci…cation. That is, apart from unobserved heterogeneity which can be handle with various econometric speci…cations, price stickiness distorts the reaction of the import prices to exchange rate changes and, hence, has to be taken into account rather in the estimation or in the interpretation. 4.3.2
Small exchange rate changes
At a more thinner grid, if the linearity assumption is rejected, parameters remain positive for depreciation except for the small exchange rate changes of magnitude 0 < q 1%, and negative for appreciation.23 That is, prices rise in response to a small depreciation, q 1%, but the impact is less than proportional. This result is in favor of incomplete ERPT both at the short and long run for the panel speci…cation. It may be of a relevant importance, and could be advanced as an explanation of the results obtained in the empirical literature, given that this category represents 34:1% of the total observations and 59:7% of the observed depreciation episodes. For the cpa classi…cation however, the linearity assumption is not rejected for small depreciation of magnitude 0 < q 1%. Therefore, one can argue that the observed low value and incomplete ERPT in the empirical literature may be explained by two factors: the relative importance of small exchange rate changes and, mainly, the nominal rigidity. Once nominal rigidity taken into account, this fact vanishes. 22
As robustness check, we run the Quadratic model speci…ed by Buissière (2013) that permits to formally test and identify the shape of the import price reaction function. We uncover that this …nding is not solely an aggregate case and remains true at a disaggregated level (NACE2). Results are available upon request. 23 Except for the long term estimate of for large exchange rate changes of 3% > q 4% where it is only signi…cant at 10% level.
16
5
Conclusion
To conclude, the extent of exchange rate pass-through and the shape of the import price reaction function to exchange rate changes are distorted by the presence of nominal rigidity. Namely, the linearity hypothesis is veri…ed once nominal rigidity is taken into account. However, in the case where it is rejected, import price reaction function is concave rather than convex, indicating the objective of …rms to maintain market shares. Consequently, and as long as ERPT is concerned, the common belief that "prices rise faster than they fall" is a result of nominal price rigidity. Therefore, care must be taken when interpreting ERPT results in the literature, especially when using aggregate prices. As shown through a number of speci…cations adopted in this paper, ERPT may be incomplete in the short run, but with relatively high value, and complete in the long run once nominal price rigidity is taken into account. While we acknowledge that micro-data are rarely available when estimating or investigating ERPT, caution must be taken when interpreting pass-through estimates if nominal price rigidity is not accounted for. Our paper provides evidence in favor of the staggered price setting framework à la Calvo or à la Taylor commonly used in the general equilibrium model. In other words, …rms gradually pass-through exchange rate changes whenever they are able to adjust prices but the commonly observed low and incomplete value of ERPT in the litterature is explained by the relative importance of small exchange rate changes and, mainly, the presence of nominal rigidity. In terms of policy implications, given that (1) the size of import price changes is small, (2) the extent of ERPT is distorted by the nominal rigidity especially for small exchange rate changes where the impact on import prices is less than proportional, and (3) the home consumption bias is important, the role of the ERPT is considerably reduced in in‡uencing in‡ation. From a normative point of view, it seems therefore important to understand the reasons behind import price rigidity for di¤erentiated goods that are characterized by high degree of nominal price rigidity. However, for homogeneous products that are characterized by low degree of price stickiness and less home consumption bias, such as the manufacture of coke and re…ned petroleum (PETRO), pass-through of exchange rate changes will have an important impact on in‡ation.
17
References [1] M. Adolfson, S. Laseen, J. Linde, and M. Villani. Bayesian estimation of an open economy dsge model with incomplete pass-through. Journal of International Economics 72(2), 481–511 (July 2007). [2] L. J. Alvarez, E. Dhyne, M. Hoeberichts, C. Kwapil, H. L. Bihan, P. Lünnemann, F. Martins, R. Sabbatini, H. Stahl, P. Vermeulen, and J. Vilmunen. Sticky prices in the euro area: A summary of new micro-evidence. Journal of the European Economic Association 4(2-3), 575–584 (04-05 2006). [3] M. Amiti, O. Itskhoki, and J. Konings. Importers, exporters, and exchange rate disconnect. NBER Working Papers 18615, National Bureau of Economic Research, Inc (Dec. 2012). [4] R. A. Auer and R. S. Schoenle. Market structure and exchange rate pass-through. Working Papers 62, Brandeis University (September 2013). [5] J. Bailliu and E. Fuji. Exchange rate pass-through and the in‡ation environment in industrialized countries: An empirical investigation. Computing in Economics and Finance 2004 135, Society for Computational Economics (August 2004). [6] K. Barhoumi. Di¤erences in long run exchange rate pass-through into import prices in developing countries: An empirical investigation. Economic Modelling 23(6), 926–951 (December 2006). [7] N. Berman, T. Mayer, and P. Martin. How do di¤erent exporters react to exchange rate changes? The Quarterly Journal of Economics 127, 437–492 (2012). [8] C. Betts and M. B. Devereux. Exchange rate dynamics in a model of pricing-tomarket. Journal of International Economics 50(1), 215–244 (February 2000). [9] H. Bouakez and N. Rebei. Has exchange rate pass-through really declined? evidence from canada. Journal of International Economics 75(2), 249–267 (July 2008). [10] A. Burstein and G. Gopinath. International prices and exchange rates. Handbook of International Economics, 4th ed. (2013). [11] M. Bussiere. Exchange Rate Pass-through to Trade Prices: The Role of Nonlinearities and Asymmetries. Oxford Bulletin of Economics and Statistics 75(5), 731–758 (October 2013). [12] J. M. Campa and L. S. Goldberg. Exchange rate pass-through into import prices: A macro or micro phenomenon? NBER Working Papers 8934, National Bureau of Economic Research, Inc (May 2002). 18
[13] J. M. Campa, L. S. Goldberg, and J. M. Gonzalez-Minguez. Exchange rate pass-through to import prices in the euro area. Sta¤ Reports 219, Federal Reserve Bank of New York (2005). [14] J. M. Campa and J. M. Gonzalez Minguez. Di¤erences in exchange rate passthrough in the euro area. European Economic Review 50(1), 121–145 (January 2006). [15] K. Christoffel, G. Coenen, and A. Warne. The new area-wide model of the euroarea: A micro-founded open-economy model for forecasting and policy analysis. Working Paper Series 944, European Central Bank (October 2008). [16] M. B. Devereux, C. Engel, and P. E. Storgaard. Endogenous exchange rate pass-through when nominal prices are set in advance. Journal of International Economics 63(2), 263–291 (July 2004). [17] C. Engel. Expenditure switching and exchange-rate policy. In “NBER Macroeconomics Annual 2002”, vol. 17 of “NBER Chapters”, pp. 231–300. National Bureau of Economic Research, Inc (2003). [18] E. Gagnon, B. R. Mendel, and R. J. Vigfusson. The hitchhiker’s guide to missing import price changes and pass-through. International Finance Discussion Papers 1040, Board of Governors of the Federal Reserve System (January 2012). [19] J. E. Gagnon and J. Ihrig. Monetary policy and exchange rate pass-through. International Finance Discussion Papers 704, Board of Governors of the Federal Reserve System (U.S.) (2001). [20] L. S. Goldberg and C. Tille. Vehicle currency use in international trade. Journal of International Economics 76(2), 177–192 (December 2008). [21] G. Gopinath and O. Itskhoki. Frequency of price adjustment and pass-through. The Quarterly Journal of Economics 125(2), 675–727 (May 2010). [22] G. Gopinath, O. Itskhoki, and R. Rigobon. Currency choice and exchange rate pass-through. American Economic Review 100(1), 304–36 (March 2010). [23] G. Gopinath and R. Rigobon. Sticky borders. The Quarterly Journal of Economics 123(2), 531–575 (05 2008). [24] C. Gust, S. Leduc, and R. Vigfusson. Trade integration, competition, and the decline in exchange-rate pass-through. Journal of Monetary Economics 57(3), 309–324 (April 2010). [25] P. J. Klenow and O. Kryvtsov. State-dependent or time-dependent pricing: Does it matter for recent u.s. in‡ation? The Quarterly Journal of Economics 123(3), 863–904 (August 2008). 19
[26] M. Marazzi and N. Sheets. Declining exchange rate pass-through to u.s. import prices: The potential role of global factors. Journal of International Money and Finance 26(6), 924–947 (October 2007). [27] T. Monacelli. Monetary policy in a low pass-through environment. Journal of Money, Credit and Banking 37(6), 1047–1066 (December 2005). [28] E. Nakamura and J. Steinsson. Lost in transit: Product replacement bias and pricing to market. American Economic Review 102(7), 3277–3316 (December 2012). [29] M. Obstfeld and K. Rogoff. New directions for stochastic open economy models. Journal of International Economics 50(1), 117–153 (February 2000). [30] J. E. Rauch. Networks versus markets in international trade. Journal of International Economics 48(1), 7–35 (June 1999). [31] F. Smets and R. Wouters. Openness, imperfect exchange rate pass-through and monetary policy. Journal of Monetary Economics 49(5), 947–981 (July 2002). [32] P. Vermeulen, D. A. Dias, M. Dossche, E. Gautier, I. Hernando, R. Sabbatini, and H. Stahl. Price setting in the euro area: Some stylized facts from individual producer price data. Journal of Money, Credit and Banking 44(8), 1631–1650 (December 2012).
20
A
Import price statistics
A.1
Concepts and speci…cations
In order to better understand the di¤erent speci…cations adopted in the statistical analysis of the import prices in section 3, let us depict in the following table a simple picture of our database structure. That is,
Entry
Exit -
period t t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8
pm i 4:61 4:64 4:64 4:64 4:68 4:68 4:70 4:70 4:70
pm i 0:03 0:00 0:00 0:04 0:00 0:02 0:00 0:00
Ii 1( ) 1 0 0 1 0 1 0 0
1 1 1 2 3 1 2 1 2
scs 1 1
srcs
1 2 3 1 2 1 2
s 1 1 1 2 3 1 2 1 2
Q
i;fs=1g
It
8
1 1 1 0 0 1 0 1 0
i;fs=2g
It
0 0 0 1 0 0 1 0 1
i;fs=3g
It
0 0 0 0 1 0 0 0 0
I i is a price change indicator. ( ) indicates that price introduction is a price change.
is the number of months elapsed since the last price change.
scs indicates the length of "Complete Spell". Notice that scs = I i : . srcs indicates the length of "Right Censored Spell". s indicates the length of complete and censored spell. Q is the quote-line i;fs= g It is the price spell, complete and censored, indicator
i Denote pm it the (log) import price of an individual product i. It is a price change indicator m m that takes the value of 1 if pit 6= pit 1 and 0 otherwise. For the sake of illustration, we assume that the individual product i enters the sample at the period t and that price introduction is a price change. The latter assumption implies that Iti = 1 at the period t and simply means that we do not consider left censoring. is the number of months elapsed since the last price change. The complete spell of length scs is de…ned to be the number of months elapsed between the current observed price change and the last one. It is given by the value of conditional on observed price change, that is, j I i = 1 . In turn, the right censored spell of length srcs is de…ned to be the number of months elapsed since the last price change but where the current period does not correspond yet to the price change. It is therefore given by the value of where no price change is observed, that is, j I i = 0 .
21
The quote-line Q, de…ned to be the number of months that elapsed between the entry of the individual product i in the sample (period t) and its exit (period t + 8), is equal to 8 months in our simple example. i;fs= g Finally, It is a price spell indicator that takes the value of 1 when the number of spell s, complete and censored, corresponds to a given value of , and 0 otherwise. It permits to classify observations by the number of spell.
A.2
Hazard rate
The expression of the conditional hazard rate in equation (1) in section 3.1 is obtained P i;fs= g as follows. The term in the denominator computes the number of observat It tions/occurrences for which the length of spell s of an individual price i last months. Namely, it corresponds to the total number of complete and censored spell of length months. In our P P i i;fs= g i;1 example, it is given by t+8 T =t IT and is equal to 5 for s = 1. While the term t It :It in the numerator computes the number of observations/occurrences having a spell of length months with price changes. Namely, it corresponds number of complete spell Pto thei total i;1 :I and is equal to 2 for s = 1. of length months. In our example, it is given by t+8 I T =t T T Finally, taking the weighted mean across individual products i yields the expression of the hazard rate conditional on a given price spell.
B B.1
Pass-through and price stickiness Speci…cation t t+1 t+2 t+3 t+4 .. .
pm 1 x o o x x .. .
pm 2 x o o o x .. .
pm 3 x x o x o .. .
pm 4 x x x o o .. .
q qt qt+1 qt+2 qt+3 qt+4 .. .
- x indicates price change - o indicates no price change Denote pm i the (log) import price of an individual product i and q the (log) real exchange rate. Assume that all products enter the sample at period t for simplicity. Under the assumption that a price introduction is a price change, the row corresponding to period t contains "x" which indicates a price change. This assumption simply means that we do not consider left censoring.
22
B.2
Speci…cation of ERPT equation conditional on price adjustment (cpa)
Concerning the "conditional on price adjustment" (cpa) speci…cation, the variables used in the estimation are constructed as follows. Let us take the case of the import price of the …rst individual good pm 1 . For this good, we observe 2 complete price spells: from t to t + 3 and t + 3 to t + 4. The …rst spell lasts 3 months and the second spell 1 month. Therefore, import m m price changes conditional on price adjustment is given by 3 pm pm 1 p1 = 1 = p1t+3 1t and m m m p1t+4 p1t+3 . De‡ating with the PPI index yields the dependent variable s p pppi in equation (4). Independent variables in turn are obtained as the cumulative sum of real exchange rate changes during each complete spell. That is in the case of individual good 1, r3 q = (qt+1 qt ) + (qt+2 qt+1 ) + (qt+3 qt+2 ) = (qt+3 qt ) and r1 q = (qt+4 qt+3 ). This yields the cumulative real exchange rate changes rs q for complete spell of length s (or more precisely, of length scs ) in equation (4).
B.3
Speci…cation of ERPT equation conditional on price change (cpc)
In turn, "conditional on price change" (cpc) speci…cation considers, for each period, only products that have a price change. That is, period t import price changes pm t in equation (5) is obtained by taking the mean of import prices that have changed at period t across m individual products. In our example, if we denote pm pm it = pit it 1 , conditional on price m m m change dependent variable pt is given by: pt+1 = mean( p3t+1 ; pm pm pm 4t+1 ), t+2 = 4t+2 , m m m pt+3 = mean( p1t+3 ; p3t+3 ) and so on. Independent variables in turn correspond simply to one period change of the real exchange rate qt and the last six lagged values.
B.4
The Threshold model
We assess non-linearity and asymmetry issues using the following Threshold model: pm t
(panel) :
pppi t
=
6 X
j
qt
j
+
j=0
(cpa) :
s 6 X
pm t c x jD
= pppi t
qs;t
s j
6 X
jD
x
qt
j
+ d0w + "t
j=0
c 0 rs qt
+
6 X j=1
c j
qs;t
s j
+
c x 0 D rs qt
+
(6)
+ d0w + "t
j=1
where Dx is a dummy variable that takes the value of 1 if x is true and 0 otherwise. We consider di¤erent hypothesis for x, namely positive vs negative exchange rate changes, and various amplitudes for exchange rate changes. Thus, for the panel and cpa speci…cation, we make a general distinction between appreciation and depreciation in order to assess asymmetry issue, and add a thinner grid of exchange rate changes in order to assess non-linearity issue. 23
The interest lies in P estimating P and testing the sign of the threshold parameters 0 or c0 6 for short term and j=0 j or 6j=0 cj for long term. We denote henceforth this set of Threshold parameters. Positive signi…cant estimate of the Threshold parameters in Table 3 indicates a more than proportional e¤ect ( + ) of the exchange rate changes into import price changes. In turn, negative estimate of the Threshold parameters indicates a less than proportional e¤ect ( ). For the panel speci…cation, a positive signi…cant estimate of = 0:24 indicates that a depreciation ( q > 0) has a more than proportional e¤ect of 0:33 ( + = 0:09 + 0:24 = 0:33) on import price in the short run. In turn, a negative signi…cant estimate of = 0:24 indicates that an appreciation ( q < 0) has a less than proportional e¤ect of 0:09 ( = 0:33 0:24 = 0:09) on import price in the short run. For the conditional on price adjustment speci…cation, a depreciation has a less than proportional e¤ect with a total impact of 0:65 (1:27 0:62 = 0:65) in the long run. In turn, an appreciation has a more than proportional e¤ect with a corresponding impact of 1: 27 (0:65 + 0:62 = 1: 27) in the long run.
C
Lag-evolution of the ERPT Figure 4: Lag-evolution of pass-through estimates Conditional on price adjustment (cpa)
Conditional on price change (cpc)
Plots of pass-through estimates for di¤erent number of lags below con…rm the choice of 6 lags included in the estimation. As shown in …gure 4, the estimated values of ERPT show a rather stable evolution for both cpa and cpc speci…cations. Moreover, long term pass-through rises gradually and reaches its maximum value within the …rst 6 months after exchange rate
24
changes. This …nding is line with Obstfeld and Rogo¤ (2000) who argued that in general, import price adjustments take place within 90 days or less.
25
D
Tables Table 1: Import price summary statistics
Group
Group DP HG RP NC
Group CFP CNFND CD IG ENER CG NC
General Quotes 38:23 [39:70]
Rigidity 6:67 [6:67]
Rauch Quotes 40:33 [38:48] 37:47 [42:29] 38:28 [36:55] 38:93 [38:41]
Rigidity 7:44 [7:14] 1:65 [2:50] 5:95 [6:31] 4:56 [5:60]
Vermeulen Quotes 39:22 [39:32] 34:55 [31:49] 46:93 [45:64] 37:12 [34:60] 44:76 [44:89] 36:97 [35:71] 45:95 [45:99]
Rigidity 4:29 [6:00] 6:06 [6:65] 9:19 [9:47] 6:18 [5:71] 1:91 [3:04] 8:60 [8:15] 8:01 [7:59]
Group MINQUA FOOD BEVE TOBA TEXT WEAP LEATH WOOD PAPE PETRO CHEM PHAR RUBB NMMP METAL FMET COMP ELEC MACH VEHI OTRA FURN OMAN EGSA
NACE2 Quotes 35:49 [35:43] 35:52 [35:24] 64:68 [59:42] 39:00 [39:00] 40:09 [39:10] 41:12 [44:39] 22:17 [16:18] 43:11 [42:83] 46:20 [44:07] 33:28 [29:63] 39:79 [36:05] 51:18 [50:92] 40:65 [35:32] 30:01 [24:45] 35:72 [38:34] 43:02 [42:56] 36:36 [40:67] 48:97 [49:09] 43:85 [42:03] 45:03 [43:71] 29:10 [26:25] 48:10 [48:46] 43:53 [38:65] 37:00 [37:00]
Rigidity 1:66 [1:97] 4:38 [5:79] 4:51 [4:98] 5:34 [5:15] 5:17 [4:74] 8:19 [12:38] 4:60 [2:83] 8:71 [7:34] 5:65 [5:86] 2:29 [4:46] 4:46 [4:59] 9:58 [10:71] 10:17 [9:54] 8:89 [7:99] 7:32 [6:50] 8:48 [6:34] 9:21 [8:38] 5:00 [4:97] 8:72 [8:76] 2:51 [3:75] 6:72 [7:78] 15:02 [14:70] 7:44 [6:38] 1:00 [1:00]
- The Table provides summary statistics on import prices based on 3 di¤erent product classi…cations which are Rauch (1999), Vermeulen et al. (2012) and NACE rév 2. - Quote-lines indicate the number of months elapsed between the entry of a given ELI in the sample and its exit. - Price rigidities indicate the average number of months where prices remain unchanged and are calculated as the inverse of the probability of price changes (frequency). - Values between square brackets are the Non-weighted mean of quote-lines or price rigidities.
26
Table 2: ERPT estimates
General Group
ST-ERPT 0:21; [0:32]; (0:60)
NACE2 LT-ERPT 0:47; [0:86z ]; (0:81z )
Group M IN Q U A FO O D
Rauch Group DP
ST-ERPT 0:19; [0:33]; (0:60)
LT-ERPT 0:39; [0:84z ]; (0:81z )
[
[ 0:72; [0:65z ]; (0:06 )
0:18; [0:30]; (0:32 ) 0:28; [0:32]; (0:83z ) 0:29; [0:28]; (0:42)
HG RP NC
BEVE
0:70; [1:07z ]; (1:14z ) 0:68; [0:82z ]; (0:54z )
TOBA TEXT W EAP L E AT H WOOD PA P E
CFP
Vermeulen ST-ERPT LT-ERPT 0:30; [0:36]; (0:59) 0:87z ; [1:35]; (1:28z )
CNFND
0:17; [0:39]; (0:86z )
0:38; [0:85z ]; (1:25z )
0:27; [0:50]; (0:46) 0:19; [0:26]; (0:56) 0:54; [0:48[ ]; (0:46) 0:22; [0:20]; (0:70) 0:23; [0:22]; (0:55)
0:85z ; [2:17]; (1:59z ) 0:42; [0:52]; (0:50) 1:2; [ 2:2]; ( 1:5[ ) 0:46; [1:10z ]; (1:17z ) 0:46; [0:95z ]; (0:74z )
Group
CD IG ENER CG NC
PETRO CHEM PHAR RUBB NMMP M E TA L FM ET COMP ELEC M ACH VEHI OTRA FURN OMAN EG SA
ST-ERPT 0:56z ; [0:46[ ]; (0:59z ) 0:24; [0:27]; (0:51) 0:23; [0:75z ]; (0:57z ) 0:22; [0:13[ ]; (2:34) 0:0[ ; [0:18[ ]; (0:03[ )
LT-ERPT 2:3; [ 3:1]; ( 2:0[ ) 0:79; [1:21z ]; (1:21z ) 0:60; [2:09z ]; (0:54) 0:55; [0:12[ ]; ( 3:3) 0:43; [1:01z ]; (0:82z )
0:10; [0:34]; (0:59) 0:66; [0:57]; (0:91z ) 0:17; [0:30]; (0:77z ) 0:35; [0:41]; (0:65z ) 0:62z ; [0:45[ ]; (0:98z ) 0:25; [0:48]; (0:75z ) 0:14; [0:28[ ]; (1:33z )
0:29; [0:63z ]; (1:15z ) 0:77z ; [1:14z ]; (0:47) 0:13[ ; [0:85z ]; (1:22z ) 0:24[ ; [0:10[ ]; (0:50z ) 1:71z ; [0:55z ]; (1:18z ) 0:97z ; [1:05z ]; (1:08z ) 0:50; [2:31z ]; (2:14)
0:11; [0:13[ ]; (0:36) 0:33; [0:32]; (1:00z ) 0:37; [0:55]; (0:35[ ) 0:19; [ 0:0[ ]; (0:83z ) 0:20; [0:32]; (0:71z ) 0:29; [0:71]; (0:46) 0:13; [0:14]; (0:48) 0:05[ ; [ 0:0[ ]; (0:05[ ) 0:04[ ; [0:03[ ]; (0:05[ ) 0:16; [0:51]; ( 0:01) 0:45; [0:33]; (1:29z ) 0:3[ ; [ 0:3[ ]; ( 0:3[ )
0:07[ ; [0:72z ]; (0:15 ) 0:35; [0:71z ]; (0:72) 0:29[ ; [ 0:0[ ]; ( 1:2[ ) 0:36; [0:79z ]; (1:17z ) 0:28; [0:77z ]; (0:71z ) 1:12z ; [3:04]; (1:58z ) 0:35; [0:96z ]; (0:88z ) 0:14[ ; [0:23[ ]; (0:21[ ) 0:07[ ; [1:16z ]; (0:49) 0:93z ; [1:55z ]; (2:18z ) 0:53; [0:95z ]; (1:61) 2:0; [ 2:0]; ( 2:0)
- The table provides results from …xed e¤ects panel estimation of ERPT equation. Three di¤erent speci…cations are considered: 1- the benchmark panel speci…cation (equation (3)) where results are depicted in the …rst value of the table, 2- the conditional on price adjustment (equation (4)) where results are depicted between square brackets, 3- the conditional on price change (equation (5)) where results are depicted between parenthesis. - ([) ERPT is not statistically di¤erent from 0 at 95% level (zero ERPT). - (z) ERPT is not statistically di¤erent from 1 at 95% level (complete ERPT).
27
[
Table 3: Threshold model
I- Panel speci…cation Exchange rate changes Threshold Percentage relative to: Sign Short term long term positive negative total q>0 0:24 57:2% q 0 0:62 48:9% rs q < 0 0:62 51:1% Sign and magnitude Short term long term positive negative total 5% < rs q 0:32 19:5% 9:5% 4% < rs q 5% 0:25 5:9% 2:9% 3% < rs q 4% 0:40 10:2% 5:0% 2% < rs q 3% 0:48 1:54 9:1% 4:4% 1% < rs q 2% 0:34 14:9% 7:3% 0% < rs q 1% 40:5% 19:8% 1% rs q < 0% 0:79 35:2% 18:0% 2% rs q < 1% 0:20 25:0% 12:8% 3% rs q < 2% 0:30 1:13 15:9% 8:1% 4% rs q < 3% 10:2% 5:2% 5% rs q < 4% 1:06 6:1% 3:1% rs q < 5% 7:6% 3:9% -
-
(
) Parameter estimate is statistically signi…cant at 5% (10%) level
q is the one period exchange rate changes rs q is the cumulative sum of exchange rate changes during the complete spell s
- Positive sign in the Threshold model indicates a more than proportional e¤ect. - Negative sign in the Threshold model indicates a less than proportional e¤ect.
28
